Document 11952120

advertisement
WP98-03
APRIL 1998
.
Working Paper
Department of Agricultural, Resource, and Managerial Economics
Cornell University, Ithaca, New York 14853-7801 USA
THE EMPIRICAL IMPACT OF BOVINE SOMATOTROPIN
ON A GROUP OF NEW YORK DAIRY FARMS
Zdenko Stefanides and Loren W. Tauer
•
.
'
It is the policy of Cornell University actively to support equality
of educational and employment opportunity. No person shall be
denied admission to any educational program or activity or be
denied employment on the basis of any legally prohibited dis­
crimination involving, but not limited to, such factors as race,
color, creed, religion, national or ethnic origin, sex, age or
handicap. The University is committed to the maintenance of
affirmative action programs which will assure the continuation
of such equality of opportunity.
"•
l
THE EMPIRICAL IMPACT OF BOVINE SOMATOTROPIN ON A GROUP OF
NEW YORK DAIRY FARMS
Zdenko Stefanides and Loren W. Tauer*
ABSTRACT
Data from a panel of New York Dairy farms were used to estimate rbST adoption
functions, and to measure the impact of rbST on milk output
and profitability per cow.
Adoption results are consistent with previous rbST adoption studies. Farm size, productivity and
education of the principal operator are the most important explanatory variables influencing
adoption. The use of rbST was found to significantly increase milk output per cow net of other
explanatory variables, correcting for self-selection with respect to rbST use. The impact on
profits, was, however, not statistically different from zero at any conventional statistical
significance level.
*Zdenko Stefanides is a graduate student and Loren W. Tauer is a professor in the Department of
Agricultural, Resource, and Managerial Economics, Cornell University. Paper presented at the
annual meeting of the American Agricultural Economics Association in Salt Lake City, Utah,
August 3-5, 1998. The authors thank Tim Mount for his assistance. Funding for Stefanides was
provided by the Andrew W. Mellon Foundation.
­
THE EMPIRICAL IMPACT OF BOVINE SOMATOTROPIN ON A GROUP OF
NEW YORK DAIRY FARMS
Recombinant Bovine Somatotropin (rbST), a synthetic version of a naturally-occurring
bovine growth hormone, is one of the first commercial agricultural technologies from
recombinant DNA technology research. In numerous experimental trials, rbST increased milk
production by 2.5 to 30 percent depending on dairy management practices (Jarvis). The most
productive herds treated with rbST were projected to see an increase in profit up to $200 per cow
each year. The question of whether such profit increases can be attained on operating farms is yet
to be answered. Tauer and Knoblauch analyzed profitability changes brought from rbST use for a
group of New York dairy farms during the first year of its availability (1994). While their study
found a significant milk production increase for farms using rbST, the impact on profit, although
substantial and positive, was not statistically different from zero. This article extends their
analysis and examines rbST impact on milk production and profitability during the first two years
of commercial availability of the product.
A panel data set of 211 NY dairy farms participating in the New York Dairy Farm
Business Summary (NYDFBS) program for the years 1993-1995 was used in the analysis. The
data provide information about pre-rbST behavior of the farms (1993) and two years of rbST
experience (1994-1995). Apart from assessing whether or not rbST has been profitably used on
these farms, this study identifies the socioeconomic characteristics of the farmers related to their
adoption decisions. Such analysis serves a dual purpose here. First, the predictions from the rbST
•
adoption models are used as a means for correcting the self-selection bias inherent in estimating
1
rbST production and profitability impacts based upon farmers themselves deciding whether or
not to use the product. Second, the adoption predictions from this ex-post study can serve as a
means of evaluating the efficacy of numerous ex-ante rbST adoption predictions (see for example
Barham, 1995).
Ex-Ante Research
Since rbST has been commercially available only since 1994, published adoption studies
have been ex-ante in nature. Most studies used a producer survey, which asks farmers whether or
not, and to what extent they plan to adopt the new technology (Lesser, Magrath and Kalter;
Zepeda; Kinnucan et aI., Saha, Love and Schwart; Klotz, Saha and Butler). The primary purpose
of these studies was to identify the socioeconomic characteristics of farmers and relate these to
their adoption intentions. The data were then used to predict aggregate adoption levels and to
assess potential rbST impacts. The predicted aggregate adoption rates range from 8 to 41 percent
for early adopters, and from 33 to 92 percent for eventual adopters (Caswell, Fuglie, and Klotz).
Predicting ex-ante expected profits required assumptions about the effects of rbST on
input use, yields, costs and the milk price. One of the first studies on rbST profitability by Fallert
et aI. predicted a $157 profit per cow per year from rbST use at a milk production increase of 8.4
lbs/day. Schmidt's estimate ofrbST profit at a milk production base of 13,500 lbs and an rbST
production response rate of 10 percent was negative $2. At 20,000 lbs of milk and a 15 percent
rbST response rate, his profit estimates ranged from $83 to $163 depending upon the price of
rbST and other input costs. Butler's estimate of net revenues from rbST also ranged from
negative values on poorly managed farms with low production, to almost $250 per cow on farms
2
III
with a base production of 20,000 lbs and an 18 percent response rate. Marion and Wills predicted
a $10 rbST profit for a 12 percent response rate at 16,000 lbs base production. Jarvis re-estimated
the model by Marion and Wills using different price and rbST response assumptions and came up
with a $198 rbST profit estimate at a 15 percent response rate and a base production of 20,000
lbs.
Models
The rbST impact on milk production and profitability is estimated within a linear
regression framework by placing a dummy variable for rbST use among other explanatory
variables. The potential endogeneity of the rbST dummy variable, however, is acknowledged and
corrected. Given the panel nature of the data, both fixed and random effects specification of the
regression equations are examined.
The linear regression equation to be estimated is:
Yit = Xit~ + OR it + eit
(1)
where Yit is a milk output or profit variable, Xit is a vector of explanatory variables, Rit is a
dummy variable for rbST use (RiFl if rbST is used, 0 otherwise), and eit is a random disturbance
assumed to be normally distributed.
If () is to measure the impact of rbST on the output or profitability of a representative
farm, farmers should be randomly assigned whether or not to use rbST. However, since farmers
themselves decide whether or not to adopt rbST this assignment is by self-selection. As
•
suggested by the rbST adoption literature the typical farmer who chooses to adopt rbST will
3
,.
likely have relatively high milk output and profit per cow whether or not rbST is used. It follows
that the dummy variable R cannot be treated as exogenous. If equation (l) is estimated by
Ordinary Least Squares (OLS) inconsistent estimates of the parameters will result.! Correction
for this self-selection bias is usually done by including an additional equation explaining the
sample selection. In our case this is an adoption decision equation relating the farmers' adoption
decision to their individual characteristics, as well as some production features of their farms.
The predictions from the adoption equation serve as instrumental variables for the rbST use
variable R in equation (1).
The adoption equation in this. study is a binary probit model. The model is put in the
following latent regression framework:
where Rit* is an unobserved index variable, 4t represents explanatory variables, and Uit is an error
term. The observed dummy variable is the farmer's decision to adopt (RiFl) or not adopt (RiFO),
where RiF 1 if Rit*>0 and RiFO if Rit* ::;;0. The error term Uit is assumed to be normally distributed
with zero mean and variance equal to one. The probability of adoption is: P(RiFl)= P(Rit*>O)=
P(4t Y+Uit>O)= P(Uit<4t Y)= c1>(4t y), where c1> is the cumulative distribution function of the
standard normal. Estimation of this model is based on the method of maximum likelihood (see
Greene or Maddala).
Given the binary probit adoption model the rbST bias in equation (1) is:
4
•
E[Yit ]
= E[Yit IR it =1]P(R it =1)
+ E[Yit IR it
=O]P(R it =0)
where use has been made of the definition of incidentally truncated bivariate normal distribution
(see Greene, p.707). It follows that upon obtaining the estimates of <I>(~tY) from the binary probit
model one can use these estimated probabilities of rbST adoption as the instrumental variable for
R it in equations (1) to correct for the self-selection bias.
Data
The data come from 211 farms that participated in the New York Dairy Farm Business
Summary (NYDFBS) for the years 1993 through 1995. The NYDFBS extension program is
primarily meant to assist dairy farmers by analyzing their business and financial records. These
farm data are also used in dairy economics research.
The farms in the program are larger than the average New York dairy farm. Farms
participating in the program in 1995 had an average herd size of 160 cows, 20,269 pounds of
•
milk were sold per cow, and the net farm income excluding appreciation averaged $50,593 per
5
farm. This compares to a NY state average of 70 cows and 16,562 pounds of milk per cow (NY
Agricultural Statistics).
It is clear that the data are not representative of New York dairy farms. They may be
representative of the better managed farms that many believe are necessary to use rbST
successfully (Patton and Heald). Extending any conclusions outside this group or to farmers in
other states would be unsound.
Recombinant bST became commercially available during February of 1994. The DFBS
surveys for 1994 and 1995 asked farmers to indicate their use of rbST in one of the five
categories as (0) not used at all, (I) stopped using in 1994 (1995 respectively), (2) used on less
than 25 percent of the herd, (3) used on 25-75 percent of the herd, (4) used on more than 75
percent of the herd. This rbST use coding has limited information content. Neither age nor
production level of individually treated cows are known. Although most of these farms are
DHIA (Dairy Herd Improvement Association) members, that organization does not code rbST
use on individual cow records. This lack of detailed rbST management information precludes
analysis on rbST use tactics, which may be complex and unique by farm. We simply infer that
any farmer using rbST believes that it is profitable on his farm. As such, the farms were simply
sorted into rbST users and non-users. Farms using rbST on some proportion of their herds during
the whole year were labeled as users (i.e., the categories 2-4). Farms which either did not use
rbST at all or stopped using it were labeled as non-users. Table 1 provides a two-way
classification of the farms sorted in this way.
Profit is defined as milk receipts minus the operating cost of producing milk. The
operating cost of producing milk only is constructed by subtracting non-milk receipts (cull cows,
6
.
calves, excess feed sold) from the total accrual operating expenses including expansion livestock.
This procedure assumes that the cost of producing non-milk products is equal to their value. Such
an approximation to estimating non-milk operating expenses can be justified by noting that the
value of non-milk products can not exceed 10 percent of the milk receipts for the farmer to be
included in the NYDFBS final data set (Smith, Knoblauch, and Putnam). Milk production per
cow is the average milk sold per cow. As a herd average, it also includes milk from cows not
treated with rbST.
Other used data in the NYDFBS survey are: herd size, milking system, number of
milkings per day, age and education of the principal operator of the farm. Farm size is considered
a surrogate for other advanced technology use (Feder, Just and Zilberman) and is measured here
as the average number of cows on the farm (COWS). The milking system (MaKSYS) used on
the farm can also be associated with production and profit differences among different farms. In
the analysis it is coded as a dummy variable equal to I if the milking system is a parlor, 0 if a
stanchion system is used (bucket and carry, dumping station, pipeline). The number of milkings
per day (TIMES) is also coded as a dummy variable equal to I if the farm milks more than twice
a day, 0 if it milks twice a day. Milk price is calculated implicitly for each farm as milk receipts
divided by pounds of milk sold. Ex ante adoption research has shown age and education to
influence rbST adoption. Education, but not age, is hypothesized to influence milk production
and profits per cow. Education is coded as a dummy variable equal to I if the principal operator
of the farm has more than a high school education, 0 otherwise.
To capture the effect of learning-by-doing, an experience variable is included among the
set of explanatory variables. An ideal experience variable would be constructed as the
7
•
accumulated product of average number of cows on the farm and the average proportion of them
treated with rbST prior to the analyzed year. In this study, however, the experience variable is
simply a 1995 dummy variable indicating whether a farmer used or did not use rbST in 1994.
Results
Adoption Function Estimates
Besides the binary probit model, an ordered probit model and a censored regression
model were also estimated to explain adoption behavior of the farmers. The results were similar
to the binary probit analysis so only the results of that simpler model are reported. The
explanatory variables for the 1994 adoption equation come from 1993, and those explaining the
1995 adoption decision are from 1994. The 1995 data were also split into groups of 1994 rbST
non-users and 1994 rbST users to determine if the second-year adoption decisions were different
given the first-year decision. Because the likelihood ratio test statistics for the equality of the
1994 binary probit model, and the model explaining the 1995 adoption behavior of 1994 non­
users was 0.8, the null hypothesis of equality of the two models was not rejected at the 5 percent
significance level. This result effectively defines only two sub-samples for the adoption model:
one is the pooled sample of previous rbST non-users, which includes all farms in 1994, and 1994
non-users in 1995. The other sample studies 1995 adoption behavior of the group of 1994 users.
In general, the results from the binary probit adoption functions shown in Table 2 are
consistent with other studies' findings. The larger (number of cows) and more productive (milk
production per cow) the farm, the greater the probability of rbST adoption. Farms using a parlor
•
type of milking system are also more likely to adopt rbST. The negative coefficient for age
..
8
suggests that younger farmers may be more likely to adopt rbST than older farmers but that effect
is not statistically significant. Farmers with more than a high school education are more likely to
adopt rbST. The negative coefficient for a 1995 year dummy (YEAR95) suggests that farms not
using rbST in 1994 are, on average, less likely to use it in 1995 than the group of all farms in
1994. If a farmer did not use rbST in 1994, he probably will not use it in 1995.
The marginal effects (slopes) for the binary probit model represent the expected change in
probability of adoption as the explanatory variable is increased by one unit. For example, if a
farm has 10 more cows than the average (131 cows for the pooled sample) and otherwise all
characteristics of the average farm, one would expect the probability of this farm to adopt rbST
will increase by about 1.2 percent compared to the average farm. The slopes for the dummy
variables (EDUC, MILKSYS) reflect the change in probability of adoption as the dummy value
changes from 0 to 1.
A comparison of actual and predicted adoption for the three models summarized in Table
2 are shown in Table 3. Prediction was good but not exceptional. For the 1994 adoption function,
152 of the 211 farms were predicted correctly as users or non-users of rbST. The prediction for
1995 for 1994 non-users was better, probably because most adoption decisions appear to have
been made in 1994. The pooled sample of previous non-users predicted rbST use or non-use
correctly in 77 percent of the cases.
Milk and Profit Equation Estimates
Coefficient estimates of the milk production per cow regression equations with fixed
effects and a binary rbST use variable are reported in Table 4. The estimates listed under the
heading "adoption exogenous" are the estimates of equation (1) alone. i.e., these results are
9
•
conditional upon farmers' decision with respect to rbST use, and thus potentially subject to self­
selection bias. The estimates of "endogenous adoption" are the estimates of the systems of
equations (1) and (2) and are corrected for self-selection bias. Random effects specifications of
the same models were estimated but rejected by the Hausman test at the 5 percent significance
level, so only the fixed effects model estimates are presented. The dependent variable in the milk
equation is cwt. of milk per cow. This is the herd average and thus includes both rbST treated
and non-treated cows.
The coefficient for the milk price variable in the milk equation had an illogical negative
sign similar to the study of Tauer and Knoblauch. Although there are plausible theoretical
explanations for an output price having a negative sign in the short-run (Tauer and Kaiser), in
this study, we opted to drop the milk price variable from the model. The milk price in our data
set is a realized price and is not necessarily the same as the expected price which farmers use for
decision making. We assume that farm and time dummies capture the information about the
expected milk price more adequately than the imputed realized milk price available in the studied
data set.
The estimated coefficients for BSTUSE suggest that the use of rbST indeed increased
milk production per cow on these farms even when controlling for other explanatory variables
and farm and time specific effects. Farms which used rbST on some portions of their herds
during the whole year saw on average their herd average milk per cow increase by about 1000
lbs. a year compared to the farms which did not use or stopped using rbST. Replacing the rbST
variable with the predictions from the adoption model to correct for self-selection bias increased
•
the corrected BSTUSE coefficient slightly in value from 10.0 to 11.3, implying there was,
"
10
contrary to a priori expectation, a negative self-selection bias in the milk equation. Tauer and
Knoblauch, who did not correct for self-selection bias, estimated a herd milk increase of 1125
lbs. for rbST users in the first year of rbST availability.
Table 5 presents analogous estimates of the profit equation. The dependent variable in
these equations is dollars of milk receipts over operating costs per cow. The estimated
coefficients for the BSTUSE variable in the profit equations are negative and not statistically
different from zero at any conventional significance level. This implies that on average these
farms are not making money using rbST. Since rbST use increased milk output, the use of
additional inputs (i.e. rbST, feed, labor, power, veterinary expenses, and milk hauling) needed to
produce rbST induced milk consumed most, if not all, of this incremental milk revenue. Tauer
and Knoblauch estimated profit over variable cost to increase by $120 per cow with rbST use,
but this estimate had a t-statistic of only 1.5, implying their estimate was also not statistically
different from zero at the 5 percent significance level.
Replacement of the original BSTUSE variable by the predictions from the adoption
models decreased the numerical value of the estimated coefficient, implying there was a positive
self-selection bias in the profit equation with respect to rbST use. This coefficient is again,
however, not statistically different from zero at any conventional significance level either.
The additional explanatory variable measuring the learning-by-doing effect was
insignificant at the 5 percent significance level in both milk and profit equations. Experience
with rbST in 1994 appears to have no significant impact on either milk or profit per cow in the
second year of rbST availability.
•
11
Why are these farmers using rbST when it is not generating a profit for them? The answer
may be twofold. First, knowledge and discussion of rbST occurred for many years before it was
commercially available. As Barham (1996) discussed, this allowed farmers to assess rbST
technology well before it was available to them. When rbST did become available, these farmers
had their adoption decision made. This is reflected in our data showing few new rbST users in
1995 that were not using rbST in 1994. This contrasts to the normal technology release when
only a few farmers first adopt and essentially assess the technology for their neighbors. Ex-post
assessment of early adopters' experiences did not occur. Secondly, it is clear that rbST increases
milk production. With this pronounced output effect, it may be difficult to assess whether it
generates profit given the myriad of various inputs that are needed for this additional milk.
Summary and Conclusions
Data from 211 New York dairy farms were used to estimate ex-post rbST adoption
functions and to measure the impact of rbST on milk output and profitability of those farms. In
general, the adoption results are consistent with other studies' findings. Farm size, productivity,
and education of the principal operator were found to positively influence the probability of
adoption.
rbST use was found to significantly increase milk production per cow even when
allowing for other explanatory variables, and farm and time specific effects. The impact on
profits was, however, insignificant as the rbST coefficient was negative and statistically not
different from zero at any conventional significance level. Correction for the self-selection by
12
•
replacing the original rbST use variables with the predictions from the adoption models did not
dramatically change these results.
The use of rbST was not profitable on average for these farms. As with all new
technologies, a learning phase is needed for farmers to understand how to make optimal use of
rbST. Perhaps two years is simply too short a time period for a thorough understanding of the
new technology and farmers are still learning how to successfully use rbST.
•
,.
13
References
Barham, B.L. "Adoption of a Politicized Technology bST and Wisconsin Dairy Farmers" Amer.
J. Agri. Econ. 78(November 1996):224-234.
Barham, B.L., "Advancing Ex Ante Adoption Studies with Ex Post Analysis: Lessons from the
bST Experience", Department of Agricultural Economics, University of Wisconsin­
Madison, August 1995.
Butler, L. J. "Economic Evaluation of BST for Onfarm Use", Bovine Somatotropin and
Emerging Issues: An Assessment, Milton C. Hallberg, ed., Boulder, CO: Westview
Press, 1992.
Caswell, M., K. Fuglie, and C. Klotz. "Agricultural Biotechnology: An Economic Perspective".
Washington, DC: US Department of Agriculture, ERS Agricultural Economic Report
No. 687,1994.
Fallert R., T. McGuckin, C. Betts, and G.Bruner. "bST and the Dairy Industry: A National,
Regional, and Farm-level Analysis", Washington, DC: US Department of Agriculture,
ERS Agricultural Economic Report No. 579, 1987.
Feder, G., R.E. Just, and D. Zilberrnan. "Adoption of Agricultural Innovations in Developing
Countries: A Survey." Econ. Develop. and Cultur. Change 33(January 1985):255-98.
Greene, W. Econometric Analysis. Englewood Cliffs, NJ: Prentice-Hall, 2nd Ed., 1993
Jarvis, L.S. The Potential Effect of Two New Biotechnologies on the World Dairy Industry.
Boulder, CO: Westview Press, 1996
•
14
Kinnucan, H., U. Hatch, J.1. Molnar, and M. Venkateswaran. "Scale Neutrality of Bovine
Somatotropin: Ex Ante Evidence from the Southeast." South. J. Agri. Econ. 8(December
1990): 1-12.
Klotz c., A. Saba, and L.1.(Bees) Butler. "The Role ofInformation in Technology Adoption: The
Case of rbST in the California Dairy Industry." Rev. Agri. Econ. 17(Sept. 1995):287-98.
Lesser, W., W. Magrath, and R. Kalter. "Projecting Adoption Rates: Application of an Ex Ante
Procedure to Biotechnology Products." N. Centro J. Agri. Econ. 8(July 1986): 159-74.
Maddala, a.s., Limited-Dependent and Qualitative Variables in Econometrics. Cambridge:
Cambridge University Press, 1983.
Marion, B.W., and R.L. Wills. "A Prospective Assessment of the Impacts of Bovine
Somatotropin: A Case Study of Wisconsin." Amer. J. Agri. Econ. 72(May 1990):326-336.
New York State Department of Agriculture and Markets. New York Agri. Statistics, 1995-1996,
Albany, NY: New York Agricultural Statistics Service, July 1996.
Patton, R. H. and C. W. Heald. "Management of BST-Supplemented Cows." Bovine
Somatotropin and Emerging Issues: An Assessment, Milton C. Hallberg, ed., Boulder,
CO: Westview Press, 1992.
Saba A., H.A. Love, and R. Schwart."Adoption of Emerging Technologies Under Output
Uncertainty." Amer. J. Agri. Econ. 76(November 1994):836-846.
Schmidt, a.H. "Economics of Using Bovine Somatotropin in Dairy Cows and Potential Impact
on the US Dairy Industry." J. Dairy Sci. 72(1989):737-745, 1989.
15
Smith, S.P., W.A.K.noblauch, and L.D.Putnam, Dairy Farm Business Summary New York State
1995, Dept. Of Agricultural, Resource, and Managerial Economics, Cornell University
Agricultural Experiment Station, August 1996.
Tauer, L.W., and H.M. Kaiser."Negative Milk Supply Response Under Constrained Profit
Maximizing Behavior." Northeast. J. Agri. Res. Eeon. l7(October 1988):11-118.
Tauer, L.W., and W.A. Knoblauch. "The Empirical Impact of Bovine Somatotropin on New
York Dairy Farms." J. Dairy Sci. 80(1997):1092-1097.
Zepeda, L. "Predicting Bovine Somatotropin Use by California Dairy Producers." West. J. Agri.
Eeon.15(July 1990):55-62.
•
16
Table 1. Two-Way Classification of Farms Sorted Into rbST Users and Non-Users
(count data)
1994
1994 Users
Non-users
1995 Non-users
96
18
1995 Users
15
82
•
17
Table 2. Binary Probit Model Estimates for rbST Adoption Function For Previous
rbST Non-users
1994 Adoption Function
(AU of 1994 Farmers)
Variable
Coeff.
Intercept
COWS
MILKCOW
AGE
EDUC
MILKSYS
-3.26
.0035
.0138
-.0074
.286
.543
Std.Error
p-value
.896
.0014
.0040
.0097
.202
.227
.000
.009
.001
.448
.156
.017
Slopes
.0014
.0055
-.0029
.114
.216
Log Likelihood =-111.9
1995 Adoption Function for 1994 Non-users
(Sample of 1141994 rbST Non-users)
Variable
Coeff.
Intercept
COWS
MILKCOW
AGE
EDUC
MILKSYS
-3.90
.0025
.0120
-.0017
.241
.839
Std.Error
p-value
1.46
.0023
.0070
.0154
.330
.358
.001
.230
.087
.914
.464
.019
Slopes
.0005
.0023
-.0032
.047
.163
Log Likelihood =-39.9
Pooled Previous Non-users Adoption Function
(All farms in 1994 and 1994 non-users in 1995)
Variable
Intercept
COWS
MILKCOW
AGE
EDUC
MILKSYS
YEAR95
Coeff.
-3.23
.0032
.0134
-.0061
.265
.629
-.693
p-value
.000
.004
.000
.455
.121
.001
.000
Std.Error
.765
.0011
.0035
.0082
.171
.190
.185
.0012
.0048
-.0022
.095
.225
-.248
..
..
Log Likelihood =-152.2
18
l.
Slopes
Table 3. Comparison of Actual and Predicted Outcomes for Previous rbST Non-users
(Predicted Outcome Has Maximum Probability)
(1 is rbST use, 0 is non-use)
1994 Adoption Function
Predicted
o
Actual
o
92 22
37 60
1
1995 Adoption Function for 1994 Non-users
Predicted
o
Actual
o
94
16
1
1
2
2
Pooled Previous Non-users Adoption Function
Predicted
Actual
o
1
o
1
185
50
25
65
•
19
Table 4. Milk Production per Cow Equation Estimates (Herd Average) Based Upon
Binary rbST Use Variable (Fixed Effects)
Adoption Endogenous··
Adoption Exogenous
Coeff.
BSTUSE·
COWS
EDUC
TIMES
MILKSYS
R2
Std. error
10.0
.01
.46
4.67
-2.16
Coeff.
1.69
.02
5.13
3.29
5.56
Std. error
11.3
.01
2.51
5.37
-3.78
.998
2.77
.02
5.26
3.36
5.67
.998
• Farm and time dummies not printed
•• rbST use dummy replaced by the predicted probabilities from the binary adoption models
IiII
20
Table 5. Profit per Cow Equation Estimates (Herd Average) Based Upon Binary rbST Use
Variable (Fixed Effects)
BSTUSE*
COWS
PRICE
EDUC
TIMES
MILKSYS
R2
Adoption Exogenous
Adoption Endogenous**
Coeff.
Coeff.
Std. error
-38.7
.63
119.3
-54.4
-54.9
8.7
55.3
.48
35.6
104.6
66.9
112.8
Std. error
-12.7
.56
120.9
-47.5
-55.8
7.7
34.4
.46
35.6
104.2
67.0
113.0
.926
.926
• Farm and time dummies not printed
** rbST use dummy replaced by the predicted probabilities from the binary adoption models
•
21
Footnotes
1 The
problem of self-selection has been addressed in a number of studies. See Maddala for a
survey. In the rbST adoption literature studies by Klotz et aI., and Saba et al. considered this
issue.
22
·E. WORKING PAPER
WPNo
•
~
Author{s)
98-02
Food Demand in China: A Case of Guangdong Province
Zhang, X., T. D. Mount and
R. Boisvert
98-01
Wilderness: Options to Preserve, Extract or Develop
Conrad, J.M.
97-25
A Comparison of Hypothetical Phone and Mail Contingent
Valuation Responses for Green Pricing Electricity
Programs
Ethier, R., G. Poe and W. Schulze
97-24
An Exploration of the Differences in National Growth:
Some Implications for Future Environmental Policy
Khanna, N., T.D. Mount and
D. Chapman
97-23
A Critical Overview of the Economic Structure of Integrated
Assessment Models of Climate Change
Khanna, N. and D. Chapman
97-22
Valuation of Groundwater Quality: Contingent Values,
Public Policy Needs, and Damage Functions
Poe,G.L.
97-21
Can Hypothetical Questions Predict Actual Participation in
Public Programs? A Field Validity Test Using a Provision
Point Mechanism
Poe, G., J. Clark and W. Schulze
97-20
Voluntary Revelation of the Demand for Public Goods
Using a Provision Point Mechanism
Rondeau, D., W.O. Schulze and
G.L. Poe
97-19
The Bioeconomics of Marine Sanctuaries
Conrad, J.M.
97-18
Introducing Recursive Partitioning to Agricultural Credit
Scoring
Novak, M. and E.L. LaDue
97-17
Trust in Japanese Interfirm Relations: Institutional
Sanctions Matter
Hagen, J.M. and S. Choe
97-16
Effective Incentives and Chickpea Competitiveness in India
Rao, K. and S. Kyle
97-15
Can Hypothetical Questions Reveal True Values? A
Laboratory Comparison of Dichotomous Choice and
Open-Ended Contingent Values with Auction Values
Balistreri, E., G. McClelland, G. Poe
and W. Schulze
97-14
Global Hunger: The Methodologies Underlying the Official
Statistics
Poleman, T.T.
97-13
Agriculture in the Republic of Karakalpakstan and Khorezm
Oblast of Uzbekistan
Kyle, S. and P. Chabot
To order single copies of ARME publications, write to: Publications, Department of Agricultural, Resource, and Managerial Economics,
Warren Hall, Cornell University, Ithaca, NY 14853-7801. Visit our Web site at http://www.cals.comell.edu/deptlarme/for a more
complete list of recent publications.
-.'
Download